How much energy can AI afford – and how reliable can it be?
In July, gAIn brought these questions to one of the largest computational science congresses in the world. Our gAIn members Gitta Kutyniok and Massimiliano Datres organized the mini-symposium “Reliable and Sustainable AI: From Mathematical Foundations to Next Generation AI Computing” at the 17th World Congress on Computational Mechanics and the 10th European Congress on Computational Methods in Applied Sciences and Engineering (WCCM-ECCOMAS 2026) in Munich, with contributions from all three gAIn universities.
Gitta Kutyniok gave a talk on the fundamental limitations of conventional digital computing and the potential of neuromorphic architectures and spiking neural networks for more reliable and energy-efficient AI. The session explored brain-inspired and analog computing paradigms – neuromorphic chips that mimic how neurons process signals, and analog hardware that exploits the continuous dynamics of physical systems – and the hard questions they raise: robustness to noise, the precision a given task really needs, and how to co-design models and circuits without sacrificing scalability.
Bringing machine learning researchers and hardware engineers into one room was the goal – and the stimulating discussions at the intersection of mathematics, AI and next-generation computing showed how much both sides gain from the exchange. Thank you to all speakers and participants.
Congress abstract: https://congressarchive.cimne.com/wccm_eccomas_2026/proposals/eb30e86c48f911f096d6000c29ddfc0c.pdf
Date
19th – 24th July, 2026
Location
WCCM-ECCOMAS 2026, Munich, Germany